How to Choose a Deepfake Protection Service

Choosing a deepfake protection service is a decision about who will handle some of the most sensitive material about you or your organisation. Marketing pages tend to sound alike, so it helps to ask the same specific questions of every provider. This buyer's checklist covers what to look for, what to ask and which answers should give you pause.

It is deliberately neutral. It doesn't compare named providers, and it applies to Eterna as much as to anyone else.

How to choose a deepfake protection service: the checklist

1. Verification standard. Ask how the provider decides something is a genuine deepfake of you. Look for a documented test that covers whether the content is synthetic, whether it depicts you, and whether it is harmful, rather than a single automated score.
2. Evidence handling. Ask what is recorded when content is found: source URL, timestamps, context, copies and integrity checks. Ask where evidence is stored, who can access it and how long it is kept.
3. Authorization checks. A responsible provider confirms you are the person depicted, or are authorized to act for them, before monitoring or taking action. Be wary of any service that will act on anyone's behalf without checks.
4. Human review. Ask which decisions are made by people. Automated tools are useful for finding signals, but decisions to report, escalate or tell a client something is harmful should involve human judgment.
5. Platform coverage. Ask which platforms and sources are actually monitored and which are not. No service covers the entire internet, and a clear list is more useful than a broad claim.
6. Reupload monitoring. Harmful content often returns after removal. Ask whether monitoring continues after a response and how recurrences are handled.
7. Transparency about limits. Ask what the provider cannot remove or influence. Content decisions rest with platforms, and some material is lawful even when it is unwelcome. A provider should say so plainly.
8. Confidentiality. Ask how your identity and case details are protected, who on the provider's team sees them, and whether the provider ever uses client stories in marketing.

Questions to ask on a first call

• Can you walk me through what happens from the moment a signal is found to the moment I hear about it?
• What do you need from me before you will act, and why?
• What will you do if the content turns out to be lawful commentary or parody?
• How will I know what action has been taken and what the platform said?
• What happens to my data if I stop working with you?

Warning signs

Some answers should prompt more questions. Be cautious of guarantees that everything will be removed, fixed timelines that a platform hasn't committed to, statistics with no source, a refusal to explain the verification process, or an offer to act against content without confirming your identity or authorization.

What a good engagement looks like

Before any monitoring starts, a responsible provider agrees the scope with you: who is being protected, which names, handles and images are in scope, which sources will be watched, and what the provider is authorized to do when something is found. It should be clear who your point of contact is and how findings will reach you.

During the engagement, you should receive findings that explain what was found, where, when and why it matters, with the evidence attached. When action is taken, you should see what was submitted and what the platform replied. If something cannot be removed, you should be told why, and what other options exist.

Matching the service to your situation

An individual facing a single incident may mainly need fast verification, evidence and help with reporting. A public figure or executive with ongoing exposure may need continuous monitoring, coordination with managers or legal advisers, and reupload follow-up. An organisation may need reporting that fits its security and communications processes.

Eterna's own approach is set out in how Eterna works and our verification methodology. If you're comparing options for a public figure, see public figure protection.